Elemental analysis of powders with surface-assisted thin film laser-induced breakdown spectroscopy

    Elemental analysis of powders with surface-assisted thin film laser-induced breakdown spectroscopy Ye Tian, Ching Cheung Hoi, Ronger Zheng, Qianli Ma, Yanping Chen, Nicole Delepine-Gilon, Jin Yu PII: DOI: Reference:

S0584-8547(16)30150-1 doi: 10.1016/j.sab.2016.08.016 SAB 5120

To appear in:

Spectrochimica Acta Part B: Atomic Spectroscopy

Received date: Revised date: Accepted date:

15 January 2016 12 August 2016 13 August 2016

Please cite this article as: Ye Tian, Ching Cheung Hoi, Ronger Zheng, Qianli Ma, Yanping Chen, Nicole Delepine-Gilon, Jin Yu, Elemental analysis of powders with surfaceassisted thin film laser-induced breakdown spectroscopy, Spectrochimica Acta Part B: Atomic Spectroscopy (2016), doi: 10.1016/j.sab.2016.08.016

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Elemental analysis of powders with surface-assisted thin

T

film laser-induced breakdown spectroscopy

IP

Ye Tiana,b, Hoi Ching Cheunga,d , Ronger Zhengb, Qianli Mae, Yanping Chene,

a

SC R

Nicole Delepine-Gilonc and Jin Yua,e,

Institut Lumière Matière, UMR5306 Université Lyon 1-CNRS, Université de Lyon, 69622 Villeurbanne Cedex, France

Optics and Optoelectronics Laboratory, Ocean University of China, 266100, Qingdao, R. P.

NU

b

China

Institut des Sciences Analytiques, UMR5280 Université Lyon 1-CNRS, Université de Lyon,

MA

c

69622 Villeurbanne Cedex, France d

Department of Physics, Hong Kong Baptist University, Waterloo Road, Kowloon,

D

Hong Kong, China e

TE

Key Laboratory for Laser Plasmas (Ministry of Education), Department of Physics and

Abstract

CE P

Astronomy, Shanghai Jiao Tong University, Shanghai 200240, China

We have developed in this work a method of elemental analysis of powdered materials with laser-induced breakdown spectroscopy (LIBS). This method requires simple sample

AC

preparation. Powders are first mixed into a 75 cSt base oil to obtain a paste which is then smeared onto the polished surface of a solid state substrate, aluminum plate for instance, in the form of a uniform thin film. The prepared sample is ablated by a high energy infrared (IR at 1064 nm) nanosecond laser pulse. The laser beam transmits through the coating layer of the material to be analyzed and induces a strong plasma from the substrate. The initial plasma interacts in turn with the coating layer, leading to the vaporization and excitation of the incorporated powder particles. The subsequent emission from the plasma includes emission lines of the elements contained in the powder, which is preferentially captured by a suitable detection system. The analysis of the recorded spectrum allows the concentration determination of the targeted elements in the powder. We first applied the method on a cellulose powder of 20 µm typical particle size. The powder was spiked with titanium dioxide 

Corresponding author. E-mail address: [email protected] (Jin Yu). 1

ACCEPTED MANUSCRIPT (TiO2) nanoparticles for Ti concentrations ranging from 25 ppm to 5000 ppm by weight. Calibration graphs were thus built to deduce figures-of-merit parameters such as the coefficient of determination (

) and the limits of detection and quantification (LoD and

T

LoQ). We optimized especially the choice of reference line for spectrum normalization, which

IP

resulted in better analytical performances. In the second step, two sets of powders, the aforementioned cellulose powder and an alumina powder with average particle size of

10

SC R

µm, were spiked with TiO2 nanoparticles. We then assessed the matrix effect between these two different powders for the determination of Ti by comparing their calibration curves. Our results show universal calibration curve in Ti determination in the two tested matrices. The

NU

results are further compared to the case where the powders were prepared in pellet.

MA

Keywords

AC

CE P

TE

D

Elemental analysis, Powders, Thin film, Normalization, Matrix effect

2

ACCEPTED MANUSCRIPT

1. Introduction

T

Elemental analysis in powdered materials represents a very important aspect of analytical

IP

needs primarily because of the fact that a very large variety of materials can be mechanically crushed and ground into powders with particle size in the range from tens to hundreds of

SC R

micrometers. Powders then can be analyzed with laser-induced breakdown spectroscopy (LIBS) as reported by numerous papers in the literature. Concerned materials include soils [13], ashes [4], geological materials [5], plants [6,7], foods [8-10], fossil or nuclear combustion

NU

materials [11-14], cosmetic powders [15], pharmaceutical products [16] and so on. Elemental analysis in powder is also needed for certain industrial processes such as glass fabrication

MA

[17]. Direct laser ablation of powders encounters nevertheless some difficulties since ejection and removal of the powder occur in the laser impact region. One of the most frequently used sample preparation for LIBS analysis of powders consists of pelletizing them. In fact, pellets

D

are very suitable for LIBS measurements with their flat and large surface. Two

TE

inconveniences however are often reported in the literature for LIBS measurements of pellets. The first one is due to the fact that powders are not all suitable for preparation by pressing

CE P

into pellets with good mechanical resistance for laser ablation, or even for sample holding. Binding materials are often needed for pellet preparation [18-20]. The second one is even severer because it corresponds to the matrix effect affecting the ability of LIBS to preform quantitative analysis with pellets [1-3,5,6,10,21]. Even though a good pellet preparation, with

AC

suitable binders if necessary, can produce pellets with similar mechanical hardness, they can behave differently against laser ablation, because of either inhomogeneity in optical absorption as well as in thermal and mechanical properties on the microscopic scale. Alternative ways are searched to prepare powder samples for LIBS measurements. One of them corresponds to loose powders retained on an adhesive tape [5,22]. It has been demonstrated that with careful sample preparation and correction for sample surface mass density of the powder retained on the tape, the same calibration curve for a given element to be quantified can be obtained for powders of various materials [22]. In this work we demonstrated and evaluated a specific method of powdered material elemental analysis with LIBS. This method has been inspired from our previous works on wear metal analysis in lubricating oils [23-25] and elemental analysis in sunscreen creams [26]. In those works, we applied oils or creams on the surface of an aluminum plate in the form of a uniform and thin coating film. A strong IR nanosecond laser pulse transmitted 3

ACCEPTED MANUSCRIPT through the coating layer and focused on the substrate, which induced a hot aluminum plasma that in turn interacted with the coating film leading to its evaporation and excitation. The result was a mixture plasma including aluminum vapor but also species from the oil or the

T

cream. A suitable detection system preferentially captured emissions from the elements

IP

contained in the thin layer. Standard spectroscopic analysis allowed therefore the determination of their concentrations in the analyzed oils or creams. Our primary concern was

SC R

efficient laser ablation of viscous liquids or soft materials like oils or creams in order to obtain high temperature plasmas which allow sensitive detection of metals in oils or creams. Our result showed that temperatures exceeding 15 000 K could be reached for the vaporized oil,

NU

which permitted efficient emissions from the constituent elements [23,24]. Our further works established the fact that the dominance of the plasma from the substrate, aluminum for

MA

instance, in the resulted mixture plasma, overcomes the matrix effect in the LIBS measurement with respect to different types of oil to be analyzed [25]. The demonstrated method, which may be generically called surface-assisted thin film LIBS analysis, is actually

D

a sensitive method free of matrix effect for elemental determination in oils. The application of

TE

this method to sunscreens demonstrated its efficiency for powder analysis, since the analyzed sunscreens were in fact prepared by mixing nanoparticles of titanium dioxide (TiO2) in a

CE P

matrix of organic cream [26].

The purpose of the present work is therefore to generalize the surface-assisted thin film LIBS analysis to elemental determination in powders with particle size in the range of several tens of micrometers. As discussed above, such generalization can finds much wider

AC

applications in important areas such as environmental pollution detection, food security, industrial process control and so on. In the following, we will first detail the sample preparation procedure, the experimental setup and the measurement protocol. We will then focus our attention on the performances of the method evaluated from the calibration graphs in terms of the determination coefficient (

) and the limits of detection and quantification

(LoD and LoQ) using a cellulose powder of 20 µm typical particle size spiked by TiO2 nanoparticles. We will especially show the efficiency of a proper choice of reference line to improve the analytical performances. In the second step, we will study the matrix effect in the determination of Ti by preparing another series of samples with an alumina powder of average particle size

10 µm, spiked with TiO2 nanoparticles. The results obtained with

surface-assisted thin film LIBS analysis will then be compared to those obtained with LIBS analysis of the corresponding pellets.

4

ACCEPTED MANUSCRIPT 2. Sample preparation, experimental setup and measurement protocol

T

2.1. Sample preparation

IP

Pure powders of cellulose [(C6H10O5)n] and alumina (Al2O3) were used as two different matrices of powdered material with clearly different compositions, organic for the first and

SC R

mineral for the second. Both of them were purchased from Sigma-Aldrich with the product number S3504 for the cellulose powder with a typical particle size of 20 µm, and 265497 for the alumina powder with an average particle size of

10 µm (purity 99.5%). The powders

NU

were spiked with pure titanium dioxide (TiO2) nanoparticles by weighing them with a microbalance. Several steps of mixing were necessary (2% then 0.2% before lower concentrations) to successively dilute the concentration of TiO2 particles into the matrix

MA

powders until the final targeted Ti concentrations in the matrix powders, in the range from 25 to 5000 ppm by weight, were obtained. Each intermediate mixture powder was put into a

D

shaking machine for vigorous shaking for 5 minutes to ensure the homogeneity of the

TE

mixtures. About twenty samples were thus prepared for each matrix powder with Ti concentrations in the desired concentration range. For surface-assisted thin film LIBS measurements, a pure aluminum plate (Techlab, Al

CE P

99.99%, Cu 0.005%, Si 0.002%, Fe 0.001%) was polished and cleaned (deionized water then alcohol). A volume of spiked powder of 0.3 ml was measured using a small teaspoon and emptied in the middle of the aluminum plate. Oil (75cSt hydrocarbon base oil from Techlab)

AC

was dropped into the powder until a volume ratio of 1:1.2 (oil:powder). The resulted material was stirred to completely mix the oil and the powder until a uniform paste was obtained. The paste was then spread using a glass slide on the surface of the aluminum plate, forming finally a uniform thin film as shown in Fig. 1. The thickness of the coating layer was estimated to be 40 ± 10 μm. In such preparation, all the materials deposited on the plate, powder and oil, were left on the surface, in such way that the volume of powders applied on the surface of the aluminum plate was kept constant for different samples. Note however that the densities of the two matrix powders cellulose and alumina were different, so the surface mass densities were different for the two series of samples. In order to normalize the recoded spectra with the surface mass density of the powders, we measured the mass densities of the two powders, and

. The obtained ratio

3.72, was then used to

correct the spectra from the spiked alumina powders in order for them to be normalized to the mass of Ti applied on the aluminum plate with the spiked cellulose powders. Note that the 5

ACCEPTED MANUSCRIPT manual preparation of thin films required technical skills. Nevertheless their uniformity and thickness could fluctuate. We will come back later in the paper to the correction of such fluctuation by normalizing the recorded spectra with a reference line. After each

T

measurement, the aluminum plate was cleaned (deionized water then alcohol) and re-polished

IP

for a new sample. In parallel, pellets were also prepared for the two series of spiked powders of cellulose and alumina. For the cellulose powders, pellets could be obtained directly by

SC R

pressing 0.5 g of powder under a pressure of 667 MPa (6.8 t/cm2) for 5 minutes. For the spiked alumina powders, pure cellulose powders were used as binder at a mass ratio of 1:4 (cellulose:alumina) to form pellets with good mechanical strength under the same pressure

NU

condition and with the same mass as for the spiked cellulose powders. The Ti concentrations of the prepared pellets were therefore recalculated by multiplying by a factor of 0.8 those of

MA

the corresponding spiked alumina powders. The resultant pellets had a diameter of 13 mm,

AC

CE P

TE

D

and a thickness of 2.5 mm and 1.5 mm for the cellulose and alumina powders, respectively.

Fig. 1. Photograph of a thin film sample on the surface of an aluminum plate with a ruler showing the dimension of the plate. The visible marks on the sample surface are caused by laser ablation.

2.2. Experimental setup and measurement protocol A detailed description of the experimental setup can be found elsewhere [24]. Briefly, a Q-switched Nd:YAG laser operated at the fundamental wavelength of 1064 nm was used for ablation, at a repetition rate of 10 Hz and with a pulse duration of 5 ns. The laser pulse energy was fixed at 120 mJ for thin film samples and 60 mJ for pellet samples. The laser beam was focused 1.5 mm below the sample surface by a lens of 50 mm focal length. The laser spot on the target surface was estimated to be 300 μm in diameter, which resulted in a fluence of 170 6

ACCEPTED MANUSCRIPT J/cm2 and 85 J/cm2 for thin films and pellets respectively. The targets were mounted on motorized micrometric displacement stages with three-dimensional displacement ability in order to provide a fresh sample surface for each laser shot or each burst of laser shots during

T

the measurement. The lens-to-sample distance was monitored to be constant during the

IP

measurement using a system combining a laser pointer in oblique incidence on the sample surface and a monitoring CCD camera. A microscope glass plate placed under the focusing

SC R

lens was used to shield the lens from the ejected debris. The plasma emission was collimated along a transverse axis perpendicular to the laser incidence direction, by a system of two lenses in fused silica with focal lengths of 75 mm and 50 mm respectively, which focused the

NU

plasma emission onto the entrance of an optical fiber of 50 μm core diameter. The captured emission was directed into an echelle spectrometer (Mechelle, Andor Technology) equipped

MA

with an intensified CCD camera (ICCD, iStar, Andor Technology). The detection system offered a spectral range from 230 to 850 nm and a spectral resolving power of

= 5000.

The ICCD camera was triggered by the signal delivered by a photodiode picking a small part

D

of the laser pulse. A mechanical shutter controlled the delivery of laser pulses to the sample.

TE

A personal computer synchronized all the events, the pulse delivery, the target displacement and the activation of the ICCD camera, during a cycle of laser ablation and emission spectrum

CE P

detection. For a complementary study of the plasma morphology, a spectroscopic imaging system [27] was also available in the experiment, along another transverse axis perpendicular to the laser beam propagation direction. For the thin film measurements, each spectrum was an accumulation of 200 single-shot

AC

ablations performed in a matrix of 5 × 40 craters, with a center-to-center distance of 800 μm between the neighboring craters (craters visible in Fig. 1). The entrance of the fiber was centered at the symmetrical axis of the plasma and situated at a height 2.0 mm above the sample surface, such distance was optimized to capture the emission from the powder. The detection window was set with a delay of 2 µs and a width of 5 μs after the impact of the laser pulse on the sample. For all the samples measured with the thin film method, the gain applied to the intensifier of the ICCD camera was kept identical in order for the measured line intensities to be comparable. For the pellet measurements, each spectrum was accumulated over 200 laser shots in 10 craters with each of them ablated by 20 consequent laser pulses, and a distance of 600 μm between the neighboring craters. Due to the different characteristics of the plasma, the distance between the entrance of the fiber and the sample surface was set to be 1.3 mm. The detection window was set with a delay of 1 µs and a width of 2 μs. The gain of the intensifier was kept identical for all the spectra recorded with pellet samples. For a 7

ACCEPTED MANUSCRIPT given sample, 6 replicate measurements were performed in order to extract the average line intensity and the standard deviation of a chosen line.

T

3. Experimental results and discussions

IP

3.1. Figures-of-merit of the method and optimization with spectrum normalization

SC R

3.1.1. Raw spectra

With the above described experimental setup and procedure, LIBS spectra were first taken with spiked cellulose powders. The most intense titanium line in the spectrum, Ti II

NU

334.9 nm line (ionization energy + excitation energy

11 eV), was used to establish the

calibration graph for quantitative determination of Ti in the cellulose powder. The use of the

MA

above resonant line of the titanium ion necessitated a check of the line shape to ensure the absence of severe self-absorption and self-reversal for the line. The raw spectra in the neighborhood of this line for spiked powders with Ti concentrations ranging from 50 to 400

D

ppm by weight are shown in Fig. 2. We can see in this figure that for the 50 ppm sample, the

TE

Ti II line can be clearly distinguished from the background noise. This is indicative of a LoD under this concentration range. We also remark that the line intensity, defined in this work for

CE P

simplicity, as the height of the line with respect to the baseline in the vicinity, generally increases with the Ti concentration. But irregularity happens. For example, we can observe in Fig. 2 a reversal in intensity change for Ti concentrations from 300 ppm to 400 ppm. This can

AC

be typically caused by a relatively important fluctuation of the thickness of the thin film on the aluminum plate. We will show in the next section how spectrum normalization with a reference line can reduce the effect of such fluctuation.

8

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

MA

Fig. 2. Raw LIBS spectra in the spectral range around the Ti II 334.9 nm line from spiked cellulose powders with different Ti concentrations ranging from 50 to 400 ppm with the thin film analysis

D

method.

TE

3.1.2. Calibration graphs, R2, LoD and LoQ: improvement with spectrum normalization To obtain the calibration graph, the intensities of the Ti II 334.9 nm line extracted from

CE P

the LIBS spectra of the spiked cellulose powders are plotted as a function of Ti concentration. Fig. 3 shows first the calibration graph built with the line intensities without any normalization [Fig. 3(a)]. The data points correspond to the average line intensities and the

AC

error bars the related standard deviations. The data are fitted with a quadratic regression, . In the figure is indicated the determination coefficient

that expresses

the degree of correlation of the experimental data to the regression [28]. The theoretical LoD is also indicated in the figure with its value extracted according to the usual definition of , where

is the standard deviation of the spectrum background and

the linear

slope of the regression curve. In order to extract the LoQ value, the confidence bands [28] are plotted in the figure (confidence level 95%). LoQ corresponds then to the highest concentration (determined by the lower confidence band) related to the lowest line intensity that can be measured with confidence over the spectrum background (determined by the higher confidence band) [29].

9

AC

CE P

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

Fig. 3. Calibration graphs for the Ti II 334.9 nm line obtained with quadratic regression (red solid curves). The experimental data points correspond to the average intensities and the error bars to the line intensity standard deviations of the replicate measurements. The confidence bands are also plotted in the figures in solid gray lines. The parameters related to the figures-of-merit of the calibration graph are indicated in the figures, their definitions and extraction methods are described in the text. The line intensities shown are without normalization (a), normalized with the C I 247.9 nm line (b), normalized with the O I 777 nm line (c), normalized with the N I 746.8 nm line (d), normalized with the H I 656.3 nm line (e) and normalized with the CN 386.1 nm band (f).

In Fig. 3, we can see that without normalization, the correlation of the data with the fitting function is not good even with a quadratic model. The

value is only 0.9355 which

is usually not satisfactory for a quantitative analytical technique [28]. Correspondingly, a mediocre LoQ of 795 ppm by weight is determined. However, a relatively low LoD value of 6 10

ACCEPTED MANUSCRIPT ppm by weight is extracted. This is because it only depends on the slope of the calibration curve at small concentrations and the background noise of the spectrum. These observations on the calibration graph of Fig. 3(a) are quite consistent with our aforementioned remarks

T

about the sample-to-sample fluctuation of the raw spectra. To compensate for such fluctuation

IP

and improve the correlation between the experimental data and the regression model, normalization of the spectra with a reference line should be used. The usual criteria for

SC R

choosing a reference line correspond to lines from a matrix element considered as an internal standard, with similar excitation energies as the line to be normalized [30]. The carbon C I 247.9 nm line was thus chosen as the reference line (excitation energy = 7.7 eV) since carbon

Fig. 3 (b). We can see

NU

is a matrix element. The calibration graph with C I 247.9 nm line normalization is shown in increased significantly to 0.9810. This value is however still not for a quantitative analytical

MA

good enough to satisfy the usual requirement of

technique [28]. The corresponding LoQ of 329 ppm by weight is also relatively high. Other attempts led us to try with emission lines of other elements available in the spectra,

D

from O, H, or N for example. The two first, O and H exist in the matrix (cellulose) but can

TE

also be contributed by the ambient gas (O2 and H2O). The third, N, can only be contributed by the ambient gas (N2). It is why we were quite surprised by the better figures-of-merit obtained

CE P

with the normalization with the lines from these elements, as can be seen in Fig. 2 (c), (e) and (d). We can see that respectively normalized with the O I 777 nm line (excitation energy = 10.8 eV), the H I 656.3 nm line (excitation energy = 12.1 eV) and the N I 746.8 nm line

AC

(excitation energy = 12.0 eV), the experimental data allow building calibration graphs with of 0.9957, 0.9922 and 0.9918, above the acceptable value for a quantitative analytical technique. Correspondingly the LoQ value is improved respectively to 246, 313 and 334 ppm weight with the normalizations with these reference lines. We can even use the band-head of the CN radical at 386.1 nm as the reference line for spectrum normalization to obtain good figures-of-merit as shown in Fig. 3 (f). We remark here that the difference between CN and C is the implication of N which is contributed by the ambient gas. We can thus deduce from the results shown in Fig. 3 that with the help of a suitable reference line from an element (or molecule), the normalized line intensities allow building calibration graphs with good figures-of-merit for quantitative analysis of Ti in cellulose powder. This proves the efficiency of the developed thin film LIBS analysis method for elemental determination in powers. The spectral normalization procedure is clearly needed because of in particular, the manual preparation of the thin film on the aluminum plate in 11

ACCEPTED MANUSCRIPT absence of any precise film homogeneity and thickness control. The most suitable reference line is not necessary provided by an internal standard defined in the usual way, by an element of the matrix. This unexpected behavior leads us to perform the investigations described in the

T

next section to better elucidate the origins of the elements providing “good reference lines”

IP

and the mechanism leading to such good fluctuation correction performance.

SC R

3.1.3. Elements from the ambient gas as good references for spectrum normalization In order to study the mechanism of the observed efficient spectrum normalization and the origin of the involved reference elements, we used a spectroscopic imaging system available in our experiment to observe the morphology of the plasmas induced in LIBS analysis of the

NU

thin film of a spiked cellulose powder. The principle and the detailed description of the used imaging system can be found elsewhere [27]. Here we only specify in Table 1 the spectral

MA

lines used for each studied species and the associated on- and off-filters. As a complementary diagnostics, we also measured the temperature profile of the plasma along the laser incidence axis, by recording the emission spectrum at different heights of the plume and by extracting

D

the electron temperature from the recorded spectra using Saha-Boltzmann plot [31] of Ti I and

TE

Ti II lines. The used lines included Ti I 453.3, 498.2, 499.1, 499.9 nm lines and Ti II 323.6, 323.9, 332.3, 336.1, 337.3 nm lines. The needed electron density for the Saha-Boltzmann plot

CE P

was extracted by the Stark broadening of the H I 656.3 nm line. The measurements were performed with a spiked cellulose powder with a Ti concentration of 5000 ppm weight. Table 1 Emission lines selected for characterizing the species of Ti II, C I, O I and CN, and the

AC

characteristics of the corresponding on- and off-filters for the spectroscopic images of the species. Species

Lines

Ti II CI OI CN

Central wavelength (nm)

Bandwidth (nm)

Filter-on

Filter-off

334.9 nm

337

340

10

247.9 nm

250

220

10

777 nm 385.1, 385.4, 386.1, 387.1 nm

780

760

10

380

370

10

The spectroscopic images for the studied species and the axial electron temperature profiles are shown in Fig. 4. We specify here that the illustrated spectroscopic images correspond to emissivity images. This means that the obtained raw emission images were treated with the Abel inversion [32]. The intensity of an emissivity images is proportional to the number density of the corresponding species in the plasma in a plane intersecting the plasma into two equal parts by passing through the symmetrical axis of the plasma, i.e. the propagation axis of the ablation laser beam. The separately obtained emissivity images for the 12

ACCEPTED MANUSCRIPT four species were then superimposed in a same picture for an easy observation of their respective positions in the plume. For a good visibility, different colors are associated to the different species in the figure: bleu to O I, white to Ti II, red to CN and green to C I. Such

T

multi-species emissivity image is shown in Fig. 4 for the different detection delays of 500,

AC

CE P

TE

D

MA

NU

SC R

IP

1000, 2000 and 4000 ns together with the corresponding axial electron temperature profile.

Fig. 4. (Color online) Multi-species emissivity images of O I (in bleu), Ti II (in white), CN (in red), and C (in green) at the delays of 500 ns (a), 1000 ns (b), 2000 ns (c) and 4000 ns (d). The physical dimension of the images is indicated in (a) and the bottom of the images corresponds to the sample surface. The axial electron temperature profiles at different delays are plotted on the right side of the images with a corresponding vertical scale.

Segregation of the different species in the plasma can be clearly observed in Fig. 4. On the one hand, in the interval when the emission spectra were recorded for LIBS measurements after a delay of 2 µs, the Ti II population is located in a zone between 1 mm and 2.5 mm away from the sample surface. While the C I population is found in a zone close to the sample surface from 0 mm to 1 mm. Combining with the observed axial temperature profile, we can

13

ACCEPTED MANUSCRIPT understand that these two populations are separated in two zones inside the ablation plume with quite different temperatures. This explains their weak correlation and the mediocre improvement of the data correlation in the calibration graph obtained by normalizing the Ti II

T

line with the C I line. On the other hand, in the same time interval, the O I population is found

IP

at the same zone in height as the Ti II population. In fact, the first encircles the second as we can see in Fig. 4. They experience therefore the same temperature range in the axial

SC R

temperature profile. As we can expect a smooth temperature variation in the transverse direction of the plasma [33], we deduce thus the two populations of Ti II and O I experience a similar environment in the plume. We understand therefore why the normalization with the O

NU

I 777 nm line leads to better figures-of-merit parameters. Not only the excitation energy of the O I line is quite close to the total excitation energy (ionization + excitation) of the Ti II 334.9

MA

nm line, but also the two species overlap each other in a same zone in height inside the plume. A similar observation can be made for the CN population in Fig. 4, which explains its good performance in the fluctuation correction of the Ti II line. Moreover, since CN radicals are

D

formed in the plume by recombination between carbon evaporated from the sample and

TE

nitrogen from the background gas, and since the C I population is found very close to the sample surface, the overlapping of the CN population with the Ti II one is necessarily due to

CE P

the nitrogen from the ambient gas.

However the question still remains to know what is the origin of the excited species observed in LIBS spectrum, such as O I and H I which can in principle come either from the matrix or the ambient gas. In order to have a clear answer, we show in Fig. 5 (a) to (d) the

AC

emissivity images of O I 777 nm line at very short delays recorded for plasmas induced with the thin film of a spiked cellulose powder. We can see that the emission from O I comes at very short delays, from the front of the plume, where is located the shockwave initiated following the material ejection during the laser ablation. Our previous works show that in the atmospheric air, excited population of O I appears due to the dissociation of O2 molecules in the shockwave [34], which occurs within a very short delay with the supersonic propagation of the shockwave, much faster than the propagation of the ablated materials. We deduce therefore that the initially observed O I population is contributed by the ambient gas. The evolution of this population [Fig. 5 (a) to (d)] clearly shows that it is supplied by a continuous contribution from the ambient gas with the further propagation of the shockwave, which becomes more spherical at longer delays. A different case corresponds to the ablation of a silicate glass (SiO2). A series of O I 777 nm line emissivity images [Fig. 5 (e) to (h)] of the plasma induced on a glass sample shows an initial population of O I developed from the front 14

ACCEPTED MANUSCRIPT of the plasma as in the case of powder thin film on aluminum plate. In addition, a second population of O I propagating upward from the sample surface appears at a delay of 90 ns and

CE P

TE

D

MA

NU

SC R

IP

T

contributes to the final total O I population in the plasma.

Fig. 5. (Color online) Emissivity images of O I (in bleu) of the plasmas induced on the aluminum plate with a thin film of powder [(a) to (d)] and on a glass sample [(e) to (h)]. The delays of image recording

AC

are indicated in the images. The physical dimension of the images is indicated in (a) and the bottom of the images corresponds to the sample surface.

We deduce therefore from the study of this section that elements contributed by the ambient gas can also provide suitable reference lines for efficient spectrum normalization which significantly improves the figures-of-merit of a calibration graph. In an inhomogeneous spectroscopic emission source like laser-induced plasma for LIBS, an additional condition should be taken into account in the choice of reference lines for spectrum normalization beside the several usual considerations. Such additional condition specifies that the reference elements should spatially overlap the population of the elements for which the emission needs to be normalized.

3.2. Study of the matrix effect: comparison with pellets

15

ACCEPTED MANUSCRIPT In this section we will focus on the matrix effect in elemental determination in powders. Especially, results will be shown for the measurements on the two series of spiked powders of cellulose and alumina. The calibration curves are built for these two powders of different

T

natures: organic powder with cellulose and mineral powder with alumina. Such comparative

IP

study will be presented for the two sample preparations: the thin film preparation developed in this work and the classic pellet preparation.

SC R

3.2.1. Raw spectra

Raw spectra from the spiked cellulose and alumina powders are shown in Fig. 6 for the two sample preparation methods of thin film and pellet. The Ti concentration of the measured

NU

samples is 400 ppm by weight. At first glance, we can already remark the similitude between the spectra of the two powders prepared in thin film, while those of the samples prepared in

AC

CE P

TE

D

MA

pellet appear very different.

Fig. 6. LIBS spectra in the spectral range from 240 nm to 420 nm of cellulose [(a), (c)] and alumina [(b), (d)] powders spiked into a Ti concentration of 400 ppm by weight, prepared in thin film [(a), (b)] or in pellet [(c), (d)].

3.2.2. Calibration curves for Ti determination in cellulose and alumina powders prepared in pellet

16

ACCEPTED MANUSCRIPT To further confirm the impression gotten from the inspection on the raw spectra, calibration curves are first plotted with the two series of spiked powders prepared in pellet. For this purpose, the line intensities of the Ti II 334.9 nm line extracted from the spectra

T

recorded for these two series of pellets are first plotted as a function of Ti concentration [Fig.

IP

7(a)]. We can clearly observe the different behaviors between the two powders. At low concentrations, the slope of the calibration curve of the cellulose powder is larger than that of

SC R

the alumina powder. This indicates a matrix effect between the two powders when they are prepared in pellet for LIBS measurement. The fact that the cellulose powder has a more sensitive response should correspond to the fact that the pellets formed from the cellulose

NU

powder present a higher mechanical hardness, which helps for a more efficient laser ablation, i.e. a bigger ablated mass and a larger atomization degree, thus a denser plasma than for the

MA

pellets formed from the alumina powder. This argumentation could be confirmed by the observed curvature of the calibration curve for the cellulose pellets at high Ti concentrations, which can correspond to self-absorption of the resonant Ti II 334.9 nm line. Calibration

D

curves are then built with a weak non-resonant Ti II 457.2 nm line as shown in Fig. 7 (b). We

TE

find that the curvature of the calibration curve of the cellulose powder disappears, which leads to an even larger discrepancy between the two calibration curves due to the matrix difference

CE P

between the pellets formed from the two powders. In addition, we can extract LoDs from the calibration curves built with pellets. A LoD of 3 ppm and 5 ppm by weight are respectively found for the cellulose and the alumina powders, which is comparable to the LoD values

AC

extracted for these two powders with thin film sample preparation.

Fig. 7. Calibration curves of the Ti II 334.9 nm line (a) and the Ti II 457.2 nm line (b) for cellulose and Al2O3 powders prepared in pellet.

17

ACCEPTED MANUSCRIPT 3.2.3. Calibration curves for Ti determination in cellulose and alumina powders prepared in thin film Calibration curves are then built with the two series of spiked powders of cellulose and

T

alumina prepared in thin film. For this purpose, the line intensities of the Ti II 334.9 nm line

IP

extracted from the LIBS spectra recorded for these two series of spiked powders were first

SC R

normalized with the O I 777 nm line, because of its efficiency in the correction of sample-tosample fluctuation of the spectra. Then the relative intensities of the Ti II 334.9 nm line obtained for the alumina powder were further divided by the density ratio between the two powders

3.72. In such way the relative line intensities are

NU

normalized with the surface mass density of the analyte (Ti) deposited on the aluminum plate surface. This second normalization was necessary due to the fact that, in the thin film

MA

preparation, the quantity of the powder deposited on the aluminum plate for each thin film sample was measured by volume. The doubly normalized intensities of the Ti II 334.9 nm line are then plotted as a function of the Ti concentration in Fig. 8. We use a linear regression,

D

, to fit the data of the two powders. We can see in the figure that, contrary to what

TE

we can observe for pellets, the two sets of data obtained with thin film LIBS analysis lead to very close calibration curves. Their slopes, b, are different only by 2.3% and their intercepts,

CE P

a, are also very close. We can then merge the two sets of data. The merged data are again fitted with a linear regression as shown in Fig. 8 by a black dotted line. We can see that the merged data still have a good correlation similarly to the separated data sets for the two

AC

powders. The resulted slope is very close to the slopes for the separated data. These results show clearly the absence of matrix effect between the cellulose and the alumina powders prepared in thin film. A universal calibration curve could thus be considered for the determination of Ti in different powders, at least in cellulose and alumina powders with particle sizes similar to those concerned in this work.

18

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

MA

Fig. 8. Calibration curves with the Ti II 334.9 nm line for cellulose and alumina powders in thin film. The fitting parameters with linear regression are indicated in the figure for the cellulose data, the

TE

4. Conclusion

D

alumina data and the merged data of the two powders.

In conclusion, we have developed and evaluated in this work a method of elemental

CE P

analysis of powdered materials. This method consists in preparing powders of rather small particle size of tens of micrometers with a viscous liquid such as a 75 cSt base oil, into a paste. The paste is then applied onto the polished and clean surface of a solid substrate such as an

AC

aluminum plate, in order to form a uniform thin coating film. The sample thus prepared can be then analyzed with a standard LIBS instrument. A first important advantage of the developed method for LIBS analysis of powders with respect to the classical pellet sample preparation is the absence of the matrix effect, at least for the two tested kinds of powder, cellulose and alumina powders which represent very different kinds of organic and mineral powders. Universal calibration curve could thus be established for elemental analysis of powders. While the comparative measurements with corresponding pellets show a clear matrix effect as we can expect. The parameters of figures-of-merit extracted from the calibration graphs for the developed method show LoDs in the range of several ppm by weight, which are indeed comparable with the LoDs in LIBS analysis of the corresponding pellet samples. Moreover, the correlation of the experimental data to the regression model is greatly improved by spectrum normalization with a properly chosen reference line. Correspondingly, the LoQs extracted according to the method of the confidence bands also 19

ACCEPTED MANUSCRIPT decrease and reach, for optimized spectrum normalizations, the range of 200 ppm by weight. We especially demonstrate that the best reference lines for spectrum normalization are not necessary belonging to an internal standard element coming from the sample matrix as

T

recommended by usual considerations in spectrochemical analysis. Indeed an element

IP

contributed by the ambient gas can also provide suitable reference lines. The additional condition that should be taken into account in particular for an inhomogeneous emission

SC R

source like laser-induced plasma, is that the population of such element overlaps well that of the element for which the emission needs to be normalized in the ablation plume. Further works are still needed to generalize the conclusions that we can draw from the present work.

NU

Especially, the effect of the particle size in a powder and the influence of the simultaneous presence of a large number of elements, as it is often the case for a real powder sample as

MA

important as soils, should be studied in detail before the developed method could be applied in real-case analyses.

D

Acknowledgements

TE

The authors thank Professor Nai Ho Cheung from Hong Kong Baptist University for fruitful scientific discussions and careful reading of the manuscript. This work was supported by the

CE P

Science and Technology Commission of Shanghai Municipality (Grant No. 15142201000) and the National Natural Science Foundation of China (Grant No. 11574209). One of the authors (Y. T.) thanks the China Scholarship Council for their support. One of the authors (C.

AC

H. C.) thanks the French General Consulate in Hong Kong and Macau for their support through the Alexandre Yersin Scholarship. One of the authors (J. Y.) thanks the French Ministry of Foreign Affairs for their support through the PHC/PROCORE program.

20

ACCEPTED MANUSCRIPT References [1]

A.S. Eppler, D.A. Cremers, D.D. Hickmott, M.J. Ferris, A.C. Koskelo, Matrix effects in the detection of Pb and Ba in soils using laser-induced breakdown spectroscopy, Appl.

B. Bousquet, J.B. Sirven, L. Canioni, Towards quantitative laser-induced breakdown

IP

[2]

T

Spectrosc. 50 (1996) 1175–1181.

[3]

SC R

spectroscopy analysis of soil samples, Spectrochim. Acta Part B 62 (2007) 1582–1589. A. Segnini, A.A.P. Xavier, P.L. Otaviani-Junior, E.C. Ferreira, A.M. Watanabe, M.A. Sperança, G. Nicolodelli, P.R. Villas-Boas, P.P.A. Oliveira, D.M.B.P. Milori, Physical and chemical matrix effects in soil carbon quantification using laser-induced break-

[4]

NU

down spectroscopy, Am. J. Anal. Chem. 5 (2014) 722–729.

A. Stankova, N. Gilon, L. Dutruch, V. Kanicky, A simple LIBS method for fast

[5]

MA

quantitative analysis of fly ashes. Fuel 11 (2010) 3468-3474. J.M. Anzano, M.A. Villoria, A. Ruiz-Medina, R.J. Lasheras, Laser-induced breakdown spectroscopy for quantitative spectrochemical analysis of geological materials: effects

D. Santos Jr., L.C. Nunes, G.G.A. de Carvalho, M.D.S. Gomes, P.F. de Souza, F.D.O.

TE

[6]

D

of the matrix and simultaneous determination, Anal. Chim. Acta 575 (2006) 230–235.

Leme, L.G.C.D. Santos, F.J. Krug, Laser-induced breakdown spectroscopy for analysis

[7]

CE P

of plant materials: A review, Spectrochim. Acta B 71-72 (2012) 3–13. G.G.A. de Carvalho, D. Santos Jr., M.S. Gomes, L.C. Nunes, M.B.B. Guerra, F.J. Krug, Influence of particle size distribution on the analysis of pellets of plant materials by

[8]

AC

laser-induced breakdown spectroscopy, Spectrochim. Acta B 105 (2015) 130–135. A. Khumaeni, M. Ramli, Y. Deguchi, Y.I. Lee, N. Idris, K.H. Kurniawan, T.J. Lie, K. Kagawa, New technique for the direct analysis of food powders confined in a small hole using transversely excited atmospheric CO2 laser-induced gas plasma, Appl. Spectrosc. 62 (2008) 1344–1348. [9]

A. Khumaeni, Z.S. Lie, H. Niki, K.H. Kurniawan, E. Tjoeng, Y.I. Lee, K. Kurihara, Y. Deguchi, K. Kagawa, Direct analysis of powder samples using transversely excited atmospheric CO2 laser-induced gas plasma at 1 atm, Anal. Bioanal. Chem. 400 (2011) 3279–3287.

[10] W.Q. Lei, J. El Haddad, V. Motto-Ros, N. Gilon-Delepine, A. Stankova, Q.L. Ma, X.S. Bai, L.J. Zheng, H.P. Zeng, J. Yu, Comparative measurements of mineral elements in milk powders with laser-induced breakdown spectroscopy and inductively coupled plasma atomic emission spectroscopy, Anal. Bioanal. Chem. 400 (2011) 3303–3313. 21

ACCEPTED MANUSCRIPT [11] J.B. Sirven, A. Pailloux, Y. M'Baye, N. Coulon, T. Alpettaz and S. Gossé, Towards the determination of the geographical origin of yellow cake samples by laser-induced breakdown spectroscopy and chemometrics, J. Anal. At. Spectrom. 24 (2009) 451-459.

T

[12] E.J. Judge, J.E. Barefield II, J.M. Berg, S.M. Clegg, G.J. Havrila, V.M. Montoya, L.A.

IP

Le, L.N. Lopez, Laser-induced breakdown spectroscopy measurements of uranium and thorium powders and uranium ore, Spectrochim. Acta Part B 83 (2013) 28–36.

SC R

[13] X.W. Li, Z. Wang, Y.T. Fu, Z. Li, J.M. Liu, W.D. Ni, Application of a spectrum standardization method for carbon analysis in coal using laser-induced breakdown spectroscopy (LIBS), Appl. Spectrosc. 68 (2014) 955–962.

NU

[14] S.C. Yao, J.L. Xu, X. Dong, B. Zhang, J.P. Zheng, J.D. Lu, Optimization of laserinduced breakdown spectroscopy for coal powder analysis with different particle flow

MA

diameters, Spectrochim. Acta B 110 (2015) 146–150. [15] M.A. Gondal, M.A. Dastageer, A.A. Naqvi, A.A. Isab, Y.W. Maganda, Detection of toxic metals (lead and chromium) in talcum powder using laser induced breakdown

D

spectroscopy, Appl. Opt. 51 (2012) 7395–7401.

TE

[16] L. St-Onge, E. Kwong, M. Sabsabia, E.B. Vadas, Quantitative analysis of pharmaceutical products by laser-induced breakdown spectroscopy, Spectrochim. Acta

CE P

Part B 57 (2002) 1131–1140.

[17] H.B. Zheng, F.Y. Yueh, T. Miller, J.P. Singh, K.E. Zeigler, J.C. Marra, Analysis of plutonium oxide surrogate residue using laser-induced breakdown spectroscopy, Spectrochim. Acta Part B 63 (2008) 968–974.

AC

[18] B. Lal, H. Zheng, F.Y. Yueh, J.P. Singh, Parametric study of pellets for elemental analysis with laser-induced breakdown spectroscopy, Appl. Opt. 43 (2004) 2792–2797. [19] M.A. Gondal, T. Hussain, Z.H. Yamini, M.A. Baig, The role of various binding materials for trace elemental analysis of powdered samples using laser-induced breakdown spectroscopy, Talanta 72 (2007) 642–649. [20] L.L. Shi, Q.Y. Lin, Y.X. Duan, A novel specimen-preparing method using epoxy resin as binding material for LIBS analysis of powder samples, Talanta 144 (2015) 1370– 1376. [21] N. Gilon, J. El-Haddad, A. Stankova, W.Q. Lei, Q.L. Ma, V. Motto-Ros, J. Yu, A matrix effect and accuracy evaluation for the determination of elements in milk powder LIBS and laser ablation/ICP-OES spectrometry, Anal. Bioanal. Chem. 401 (2011) 2681–2689.

22

ACCEPTED MANUSCRIPT [22] S.I. Gornushkin, I.B. Gornushkin, J.M. Anzano, B.W. Smith, J.D. Winefordner, Effective normalization technique for correction of matrix effects in laser-induced breakdown spectroscopy detection of magnesium in powdered samples, Appl. Spectrosc.

T

56 (2002) 433–436.

IP

[23] J.S. Xiu, X.S. Bai, E. Negre, V. Motto-Ros, J. Yu, Indirect laser-induced breakdown of transparent thin gel layer for sensitive trace element detection, Appl. Phys. Lett. 102

SC R

(2013) 244101.

[24] J.S. Xiu, V. Motto-Ros, G. Panczer, R.E. Zheng, J. Yu, Feasibility of wear metal analysis in oils with ppm and sub-ppm sensitivity using laser-induced breakdown

NU

spectroscopy of thin oil layer on metallic target, Spectrochim. Acta Part B 91 (2014) 24–30.

MA

[25] L.J. Zheng, F. Cao, J.S. Xiu, X.S. Bai, V. Motto-Ros, N. Gilon, H.P. Zeng, J. Yu, On the performance of laser-induced breakdown spectroscopy for direct determination of trace metals in lubricating oils, Spectrochim. Acta Part B 99 (2014) 1–8.

D

[26] J. Menneveux, F. Wang, S. Lu, X.S. Bai, V. Motto-Ros, N. Gilon, Y.P. Chen, J.Yu,

TE

Direct determination of Ti content in sunscreens with laser-induced breakdown spectroscopy: Line selection method for high TiO2 nanoparticle concentration,

CE P

Spectrochim. Acta Part B 109 (2015) 9–15. [27] V. Motto-Ros, Q.L. Ma, S. Grégoire, W.Q. Lei, X.C. Wang, F. Pelascini, F. Surma, V. Detalle, J. Yu, Dual-wavelength differential spectroscopic imaging for diagnostics of laser-induced plasma, Spectrochim. Acta Part B 74–75 (2012) 11–17.

AC

[28] J.-M. Mermet, Calibration in atomic spectrometry: A tutorial review dealing with quality criteria, weighting procedures and possible curvatures, Spectrochim. Acta Part B 65 (2010) 509–523. [29] J.-M. Mermet, Limit of quantitation in atomic spectrometry: an unambiguous concept? Spectrochim. Acta Part B 63 (2008) 166–182. [30] N.B. Zorov, A.A. Gorbatenko, T.A. Labutin, A.M. Popov, A review of normalization techniques in analytical atomic spectrometry with laser sampling: From single to multivariate correction, Spectrochim. Acta Part B 65 (2010) 642–657. [31] J.A. Aguilera, C. Aragón, Multi-element Saha–Boltzmann and Boltzmann plots in laserinduced plasmas, Spectrochim. Acta Part B 62 (2007) 378–385. [32] Qianli Ma, Vincent Motto-Ros, Xueshi Bai and Jin Yu, Experimental investigation of the structure and the dynamics of nanosecond laser-induced plasma in one-atmosphere argon ambient gas, Appl. Phys. Lett. 103 (2013) 204101. 23

ACCEPTED MANUSCRIPT [33] Q. L. Ma, V. Motto-Ros, W. Q. Lei, M. Boueri, X. S. Bai, L. J. Zheng, H. P. Zeng, and J. Yu, Temporal and spatial dynamics of laser-induced aluminum plasma in argon background at atmospheric pressure: interplay with the ambient gas, Spectrochim. Acta

T

Part B 65 (2010) 896-907.

IP

[34] Xueshi Bai, Fan Cao, Vincent Motto-Ros, Yanping Chen and Jin Yu, Morphology and characteristics of laser-induced aluminum plasma in argon and in air: a comparative

AC

CE P

TE

D

MA

NU

SC R

study, Spectrochim. Acta B 113 (2015) 158-166.

24

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

AC

CE P

TE

D

Figure 1

25

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

AC

CE P

Figure 2

26

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

AC

CE P

Figure 3

27

AC

CE P

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

Figure 4

28

AC

CE P

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

Figure 5

29

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

AC

CE P

Figure 6

30

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

AC

CE P

Figure 7a

31

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

AC

CE P

Figure 7b

32

TE

D

MA

NU

SC R

IP

T

ACCEPTED MANUSCRIPT

AC

CE P

Figure 8

33

ACCEPTED MANUSCRIPT Table 1 Species

Lines

Ti II

Central wavelength (nm)

Bandwidth (nm)

Filter-off

334.9 nm

337

340

10

CI

247.9 nm

250

220

10

OI

777 nm 385.1, 385.4, 386.1, 387.1 nm

780

760

10

380

370

IP SC R

NU MA D TE CE P AC

CN

34

T

Filter-on

10

ACCEPTED MANUSCRIPT Highlights of the manuscript entitled “Elemental analysis in powders with surface-assisted thin film laser-induced breakdown spectroscopy” by Ye Tian et al.

AC

CE P

TE

D

MA

NU

SC R

IP

T

- Powder preparation for LIBS by mixing with base oil and by applying the obtained paste on a substrate surface. - Elemental analysis in powders with surface-assisted thin film laser-induced breakdown spectroscopy. - Spectrum normalization with reference lines of excited elements from the ambient gas. - Universal calibration curves for elemental determination in different powders.

35